Jump to content

Axiom of determinacy: Difference between revisions

From Wikipedia, the free encyclopedia
Content deleted Content added
Alternate proof of incompatibility with AC; unsure of direct source (see my section on talk soon)
Line 37: Line 37:
=== Using a choice function on a partition of the continuum into size-2 sets ===
=== Using a choice function on a partition of the continuum into size-2 sets ===


In this construction, the use of the axiom of choice is similar to the choice of "left socks" as stated in the quote by Bertrand Russell at [[Axiom_of_choice#Quotations]].
In this construction, the use of the axiom of choice is similar to the choice of socks as stated in the quote by Bertrand Russell at [[Axiom_of_choice#Quotations]].


In a ω-game, the two players are generating the sequence <math>{a(1), b(2), a(3), b(4) ...}</math>, an element in ''ω<sup>ω</sup>'', where our convention is that 0 is not a natural number, hence neither player can choose it. Define the function <math>f: \omega^\omega \rightarrow \{0, 1\}^\omega </math> such that f(r) is the unique sequence of length ω whose first term equals 0 and whose [[run-length encoding]] equals r. This can be shown to be a bijection between the domain and the subset of
In a ω-game, the two players are generating the sequence <math>{a(1), b(2), a(3), b(4) ...}</math>, an element in ''ω<sup>ω</sup>'', where our convention is that 0 is not a natural number, hence neither player can choose it. Define the function <math>f: \omega^\omega \rightarrow \{0, 1\}^\omega </math> such that f(r) is the unique sequence of length ω whose first term equals 0 and whose [[run-length encoding]] equals r. This can be shown to be a bijection between the domain and the subset of
Line 50: Line 50:
Suppose that player 2's response (always according to q) to [1 + b<sub>1</sub>] is b<sub>2</sub>, and player 2's response to [1, b<sub>1</sub>, b<sub>2</sub>] is b<sub>3</sub>. When player 1 is constructing p, they only aim to beat q, and therefore they only have to handle the response b<sub>2</sub> to their first move. Player 1 specifies that their response to [1 + b<sub>1</sub>, b<sub>2</sub>] is b<sub>3</sub>. In general, for even n, denote player 2's response to [1 + b<sub>1</sub>, ..., b<sub>n-1</sub>] by b<sub>n</sub> and player 2s response to [1, b<sub>1</sub>, ..., b<sub>n</sub>] by b<sub>n + 1</sub>. Then player 1 specifies in p that their response to [1 + b<sub>1</sub>, b<sub>2</sub>, ..., b<sub>n</sub>] is b<sub>n + 1</sub>. Strategy q is presumed to be winning, and game-result r in ''ω<sup>ω</sup>'' given by [1, b<sub>1</sub>, ...] is one possible sequence allowed by q, so r must be winning for player 2 and h(f(r)) must not be in U. Game-result r' in ''ω<sup>ω</sup>'' given by [1 + b<sub>1</sub>, b<sub>2</sub>, ...] is also a sequence allowed by q (specifically, q playing against p), so h(f(r')) must not be in U. However, f(r) and f(r') differ in all but the first term (by the nature of run-length encoding and an offset of 1), so f(r) and f(r') are in complement equivalent classes, so h(f(r)), h(f(r')) cannot both be in U, contradicting the assumption that q is a winning strategy.
Suppose that player 2's response (always according to q) to [1 + b<sub>1</sub>] is b<sub>2</sub>, and player 2's response to [1, b<sub>1</sub>, b<sub>2</sub>] is b<sub>3</sub>. When player 1 is constructing p, they only aim to beat q, and therefore they only have to handle the response b<sub>2</sub> to their first move. Player 1 specifies that their response to [1 + b<sub>1</sub>, b<sub>2</sub>] is b<sub>3</sub>. In general, for even n, denote player 2's response to [1 + b<sub>1</sub>, ..., b<sub>n-1</sub>] by b<sub>n</sub> and player 2s response to [1, b<sub>1</sub>, ..., b<sub>n</sub>] by b<sub>n + 1</sub>. Then player 1 specifies in p that their response to [1 + b<sub>1</sub>, b<sub>2</sub>, ..., b<sub>n</sub>] is b<sub>n + 1</sub>. Strategy q is presumed to be winning, and game-result r in ''ω<sup>ω</sup>'' given by [1, b<sub>1</sub>, ...] is one possible sequence allowed by q, so r must be winning for player 2 and h(f(r)) must not be in U. Game-result r' in ''ω<sup>ω</sup>'' given by [1 + b<sub>1</sub>, b<sub>2</sub>, ...] is also a sequence allowed by q (specifically, q playing against p), so h(f(r')) must not be in U. However, f(r) and f(r') differ in all but the first term (by the nature of run-length encoding and an offset of 1), so f(r) and f(r') are in complement equivalent classes, so h(f(r)), h(f(r')) cannot both be in U, contradicting the assumption that q is a winning strategy.


Similarly, suppose that p is a winning strategy for player 1; the argument is similar but now uses the fact that equivalence classes were defined by allowing an arbitrarily large finite number of terms to differ. Let a<sub>1</sub> be player 1's first move. In general, for even n, denote player 1's response to [a<sub>1</sub>, 1] (if n = 2) or [a<sub>1</sub>, 1, a<sub>2</sub>, ... a<sub>n-1</sub>] by a<sub>n</sub> and player 1's response to [a<sub>1</sub>, 1 + a<sub>2</sub>, ... a<sub>n</sub>] by a<sub>n + 1</sub>. Then game-result r given by [a<sub>1</sub>, 1, a<sub>2</sub>, a<sub>3</sub>, ...] is allowed by p so that h(f(r)) must be in U; also game-result r' given by [a<sub>1</sub>, 1 + a<sub>2</sub>, a<sub>3</sub>, ...] is also allowed by p so that h(f(r')) must be in U. However, f(r) and f(r') differ in all but the first a<sub>1</sub> + 1 terms, so they are in complement equivalent classes, so h(f(r)) and h(f(r')) cannot both be in U, contradicting that p is a winning strategy.
Similarly, suppose that p is a winning strategy for player 1; the argument is similar but now uses the fact that equivalence classes were defined by allowing an arbitrarily large finite number of terms to differ. Let a<sub>1</sub> be player 1's first move. In general, for even n, denote player 1's response to [a<sub>1</sub>, 1] (if n = 2) or [a<sub>1</sub>, 1, a<sub>2</sub>, ... a<sub>n-1</sub>] by a<sub>n</sub> and player 1's response to [a<sub>1</sub>, 1 + a<sub>2</sub>, ... a<sub>n</sub>] by a<sub>n + 1</sub>. Then game-result r given by [a<sub>1</sub>, 1, a<sub>2</sub>, a<sub>3</sub>, ...] is allowed by p so that h(f(r)) must be in U; also game-result r' given by [a<sub>1</sub>, 1 + a<sub>2</sub>, a<sub>3</sub>, ...] is also allowed by p so that h(f(r')) must be in U. However, f(r) and f(r') differ in all but the first a<sub>1</sub> + 1 terms, so they are in complement equivalent classes, so h(f(r)) and h(f(r')) cannot both be in U, contradicting that p is a winning strategy.


== Large cardinals and the axiom of determinacy ==
== Large cardinals and the axiom of determinacy ==

Revision as of 19:24, 23 August 2023

In mathematics, the axiom of determinacy (abbreviated as AD) is a possible axiom for set theory introduced by Jan Mycielski and Hugo Steinhaus in 1962. It refers to certain two-person topological games of length ω. AD states that every game of a certain type is determined; that is, one of the two players has a winning strategy.

Steinhaus and Mycielski's motivation for AD was its interesting consequences, and suggested that AD could be true in the smallest natural model L(R) of a set theory, which accepts only a weak form of the axiom of choice (AC) but contains all real and all ordinal numbers. Some consequences of AD followed from theorems proved earlier by Stefan Banach and Stanisław Mazur, and Morton Davis. Mycielski and Stanisław Świerczkowski contributed another one: AD implies that all sets of real numbers are Lebesgue measurable. Later Donald A. Martin and others proved more important consequences, especially in descriptive set theory. In 1988, John R. Steel and W. Hugh Woodin concluded a long line of research. Assuming the existence of some uncountable cardinal numbers analogous to , they proved the original conjecture of Mycielski and Steinhaus that AD is true in L(R).

Types of game that are determined

The axiom of determinacy refers to games of the following specific form: Consider a subset A of the Baire space ωω of all infinite sequences of natural numbers. Two players, I and II, alternately pick natural numbers

n0, n1, n2, n3, ...

After infinitely many moves, a sequence is generated. Player I wins the game if and only if the sequence generated is an element of A. The axiom of determinacy is the statement that all such games are determined.

Not all games require the axiom of determinacy to prove them determined. If the set A is clopen, the game is essentially a finite game, and is therefore determined. Similarly, if A is a closed set, then the game is determined. It was shown in 1975 by Donald A. Martin that games whose winning set is a Borel set are determined. It follows from the existence of sufficiently large cardinals that all games with winning set a projective set are determined (see Projective determinacy), and that AD holds in L(R).

The axiom of determinacy implies that for every subspace X of the real numbers, the Banach–Mazur game BM(X) is determined (and therefore that every set of reals has the property of Baire).

Incompatibility of the axiom of determinacy with the axiom of choice

Under assumption of the axiom of choice, we present two separate constructions of counterexamples to the axiom of determinacy. It follows that the axiom of determinacy and the axiom of choice are incompatible.

Using well-ordering of the continuum

The set S1 of all first player strategies in an ω-game G has the same cardinality as the continuum. The same is true for the set S2 of all second player strategies. Let SG be the set of all possible sequences in G, and A be the subset of sequences of SG that make the first player win. With the axiom of choice we can well order the continuum, and we can do so in such a way that any proper initial portion has lower cardinality than the continuum. We use the obtained well ordered set J to index both S1 and S2, and construct A such that it will be a counterexample.

We start with empty sets A and B. Let α J be the index of the strategies in S1 and S2. We need to consider all strategies S1 = {s1(α)} of the first player and all strategies S2 = {s2(α)} of the second player to make sure that for every strategy there is a strategy of the other player that wins against it. For every strategy of the player considered we will generate a sequence that gives the other player a win. Let t be the time whose axis has length ℵ0 and which is used during each game sequence. We create the counterexample A by transfinite recursion on α:

  1. Consider the strategy s1(α) of the first player.
  2. Apply this strategy on an ω-game, generating (together with the first player's strategy s1(α)) a sequence {a(1), b(2), a(3), b(4),...,a(t), b(t+1),...}, which does not belong to A. This is possible, because the number of choices for {b(2), b(4), b(6), ...} has the same cardinality as the continuum, which is larger than the cardinality of the proper initial portion { β J | β α } of J.
  3. Add this sequence to B (if it is not already in B), to indicate that s1(α) loses (on {b(2), b(4), b(6), ...}).
  4. Consider the strategy s2(α) of the second player.
  5. Apply this strategy on an ω-game, generating (together with the second player's strategy s2(α)) a sequence {a(1), b(2), a(3), b(4),...,a(t), b(t+1),...}, which does not belong to B. This is possible, because the number of choices for {a(1), a(3), a(5),...} has the same cardinality as the continuum, which is larger than the cardinality of the proper initial portion { β J | β α } of J.
  6. Add this sequence to A (if it is not already in A), to indicate that s2(α) loses (on {a(1), a(3), a(5), ...}).
  7. Process all possible strategies of S1 and S2 with transfinite induction on α. For all sequences that are not in A or B after that, decide arbitrarily whether they belong to A or to B. So B is the complement of A.

Once this has been done, prepare for an ω-game G. If you give me a strategy s1 of the first player, then there is an α J such that s1 = s1(α), and we constructed A such that s1(α) fails (on certain choices {b(2), b(4), b(6), ...} of the second player). Hence s1 fails. Similarly, any other strategy of either player fails.

Using a choice function on a partition of the continuum into size-2 sets

In this construction, the use of the axiom of choice is similar to the choice of socks as stated in the quote by Bertrand Russell at Axiom_of_choice#Quotations.

In a ω-game, the two players are generating the sequence , an element in ωω, where our convention is that 0 is not a natural number, hence neither player can choose it. Define the function such that f(r) is the unique sequence of length ω whose first term equals 0 and whose run-length encoding equals r. This can be shown to be a bijection between the domain and the subset of the codomain whose first term equals 0.

Observe the equivalence relation on such that two sequences are equivalent if and only if they differ in a finite number of terms. This partitions the set into equivalence classes, and let T be the set of equivalence classes (such that T has the cardinality of the continuum). Define the complement of any sequence s in to be the sequence s1 that differs in each term. Define the function such that for any sequence s in , h applied to the equivalence class of s equals the equivalence class of the complement of s (which is well-defined because if s and s' are equivalent, then their complements are equivalent. One can show that h is an involution with no fixed points, and thus we have a partition of T into size-2 subsets such that each subset is of the form {t, h(t)}. Using the axiom of choice, we can choose one element out of each subset. In other words, we are choosing "half" of the elements of T, a subset that we denote by U, such that t is in U iff h(t) is not in U.

Next, we define the subset A of ωω in which player 1 wins: A is the set of all r such that h(f(r)) is in U. We now claim that neither player has a winning strategy, using a strategy-stealing argument. Denote the current game state by a finite sequence of natural numbers (so that if the length of this sequence is even, then player 1 is next to play; otherwise player 2 is next to play).

Suppose that q is a (deterministic) winning strategy for player 2. Player 1 can construct a strategy p that beats q as follows. Suppose that player 2's response (according to q) to [1] is b1. Then player 1 specifies in p that a(1) = 1 + b1. (Roughly, player 1 is now playing as the second player in a second parallel game; player 1's winning set in the second game equals player 2's winning set in the original game, and this is a contradiction. Nevertheless, we continue more formally).

Suppose that player 2's response (always according to q) to [1 + b1] is b2, and player 2's response to [1, b1, b2] is b3. When player 1 is constructing p, they only aim to beat q, and therefore they only have to handle the response b2 to their first move. Player 1 specifies that their response to [1 + b1, b2] is b3. In general, for even n, denote player 2's response to [1 + b1, ..., bn-1] by bn and player 2s response to [1, b1, ..., bn] by bn + 1. Then player 1 specifies in p that their response to [1 + b1, b2, ..., bn] is bn + 1. Strategy q is presumed to be winning, and game-result r in ωω given by [1, b1, ...] is one possible sequence allowed by q, so r must be winning for player 2 and h(f(r)) must not be in U. Game-result r' in ωω given by [1 + b1, b2, ...] is also a sequence allowed by q (specifically, q playing against p), so h(f(r')) must not be in U. However, f(r) and f(r') differ in all but the first term (by the nature of run-length encoding and an offset of 1), so f(r) and f(r') are in complement equivalent classes, so h(f(r)), h(f(r')) cannot both be in U, contradicting the assumption that q is a winning strategy.

Similarly, suppose that p is a winning strategy for player 1; the argument is similar but now uses the fact that equivalence classes were defined by allowing an arbitrarily large finite number of terms to differ. Let a1 be player 1's first move. In general, for even n, denote player 1's response to [a1, 1] (if n = 2) or [a1, 1, a2, ... an-1] by an and player 1's response to [a1, 1 + a2, ... an] by an + 1. Then game-result r given by [a1, 1, a2, a3, ...] is allowed by p so that h(f(r)) must be in U; also game-result r' given by [a1, 1 + a2, a3, ...] is also allowed by p so that h(f(r')) must be in U. However, f(r) and f(r') differ in all but the first a1 + 1 terms, so they are in complement equivalent classes, so h(f(r)) and h(f(r')) cannot both be in U, contradicting that p is a winning strategy.

Large cardinals and the axiom of determinacy

The consistency of the axiom of determinacy is closely related to the question of the consistency of large cardinal axioms. By a theorem of Woodin, the consistency of Zermelo–Fraenkel set theory without choice (ZF) together with the axiom of determinacy is equivalent to the consistency of Zermelo–Fraenkel set theory with choice (ZFC) together with the existence of infinitely many Woodin cardinals. Since Woodin cardinals are strongly inaccessible, if AD is consistent, then so are an infinity of inaccessible cardinals.

Moreover, if to the hypothesis of an infinite set of Woodin cardinals is added the existence of a measurable cardinal larger than all of them, a very strong theory of Lebesgue measurable sets of reals emerges, as it is then provable that the axiom of determinacy is true in L(R), and therefore that every set of real numbers in L(R) is determined.

Projective ordinals

Moschovakis introduced the ordinals , which is the upper bound of the length of -norms (injections of a set into the ordinals), where is a level of the projective hierarchy. Assuming AD, all are initial ordinals, and we have , and for the th Suslin cardinal is equal to .[1]

See also

References

Inline citations

  1. ^ V. G. Kanovei, The axiom of determinacy and the modern development of descriptive set theory, UDC 510.225; 510.223, Plenum Publishing Corporation (1988) p.270,282. Accessed 20 January 2023.

Further reading